Why your word count never matches Word's
28 Sep 2026 · Text · 6 min read
You finish an essay, check the count in your writing app — 498 words, safely under the 500-word limit — then paste it into Word and it says 511. Thirteen words appear from nowhere, and the deadline is in an hour. This isn't a bug in either program. It's a disagreement about what a word is, and every piece of software that counts words has quietly picked a side. Once you see the fault lines — hyphens, dashes, numbers, apostrophes — the mystery numbers start making sense.
What is a word, anyway?
Ask a linguist and you'll get a twenty-minute answer. Ask a programmer and you'll get one line of code, and that's the problem: there are several reasonable one-line definitions, and they disagree. The simplest counter splits text on whitespace: every chunk between spaces is a word. "The cat sat" is 3 words, done. But then "well-known" arrives — one chunk, so one word — while Microsoft Word, using a far more elaborate set of rules, also calls it one word. Agreement so far. Now try "1,200": the whitespace splitter sees one chunk ("1,200"), Word sees one word, but a counter that splits on anything that isn't a letter sees two ("1" and "200"). Three definitions, two different answers, and we haven't left the first paragraph.
The honest position is that "word count" is not a measurement like weight or length. It's a convention — a counting rule somebody chose. The number is only meaningful relative to the rule, which is why the rest of this article is a tour of the rules and where they split.
The hyphen problem
Hyphenated compounds are the most common source of mismatch. Consider "state-of-the-art design":
- A whitespace-splitting counter sees one chunk: 1 word for the compound (4 words total).
- Microsoft Word's counter treats the whole hyphenated compound as 1 word too (4 total).
- A counter that splits on punctuation sees "state", "of", "the", "art": 4 words for the compound (7 total).
So the same five-word phrase scores anywhere from 4 to 7 depending on the tool. Neither is wrong in any deep sense — but if your university's submission portal uses the punctuation-splitting kind and you wrote to Word's count, you're over the limit without knowing it. Hyphens are doing genuine linguistic work (they glue modifiers together), which is why the "one word" camp has the better argument — but plenty of software never got the memo.
Dashes, and the em dash in particular
The em dash — this thing — causes a special kind of chaos because people use it three ways: with spaces ("word — word"), without ("word—word"), and inconsistently. With spaces around it, a whitespace counter sees the dash as its own token: "word — word" is 3 words. Word's counter is smarter: it knows a standalone dash isn't a word and counts 2. Without spaces, everyone agrees on 2 — "word—word" is one chunk containing no spaces, but Word still finds the two words inside it.
The en dash in ranges ("pages 12–18") is similar: one chunk to a naive counter, and Word counts the numbers around it as words. The practical lesson: if your text is full of dashes, expect the biggest gaps between counters, sometimes 2–3% of the total on dash-heavy prose.
Numbers: the quiet inflators
Numbers are where naive counters inflate. Take the sentence "Revenue grew 1,200 percent in 2024.":
| Counter style | Count | How it sees "1,200" |
|---|---|---|
| Whitespace split | 6 | One token |
| Microsoft Word | 6 | One word |
| Split on non-letters | 7 | "1" + "200" = two |
| Letters-only tokens | 4 | Not a word at all |
Currency ("$19.99"), decimals ("3.14"), percentages ("45%") and dates ("2026-10-15") all behave the same way: some counters see one word, some see two or three, some see none. A financial report dense with figures can easily differ by 5% between tools — on a 2,000-word report, that's a hundred phantom words.
Apostrophes and the contraction question
Here's a rare patch of agreement: "don't", "it's", "we've" are one word in virtually every counter, including Word. The apostrophe is treated as part of the word, which matches how we speak — nobody thinks "don't" is two words. The edge cases are possessives ending in s ("the dogs' bowls" — one word, the apostrophe just hangs there), and the rock-and-roll casualties: "rock 'n' roll" is three words to Word (the 'n' counts) but two to a whitespace counter that strips the quotes differently. Dialect and style ("y'all", "'twas") occasionally confuse simpler counters, but for normal prose, apostrophes are the least of your worries.
The languages without spaces
Everything above assumes spaces between words, which Chinese, Japanese and Thai don't use. "Word count" is meaningless for these languages in the Western sense — counters fall back to character counts, and translation billing switches to characters or "standard lines". If you work across languages, compare character counts, not word counts; it's the only number that travels.
What Word counts that you can't see
There's one more gap people discover at the worst moment: Word doesn't only count the body text. Depending on your settings, its count can include footnotes and endnotes, text inside text boxes and shapes, and even hidden text — while the online counter you pasted into sees only what you pasted. A thesis with 80 footnotes can read 300 words heavier in Word than in a paste-based tool, through no disagreement about hyphens at all. Word's status-bar count has a checkbox precisely for this ("Include textboxes, footnotes and endnotes"), which means the same document gives two different "Word counts" depending on a checkbox. When numbers diverge wildly — not by 2% but by 10% — look here first: you're probably counting different documents, not different words.
Why it matters: limits, money and grades
This isn't trivia. Word limits gate university submissions, journal articles, grant applications and writing competitions — and the venue's counter is the only one that matters. A 500-word limit checked by a strict counter will reject the essay your app scored at 498. Translators bill per source word, and a 3% counting difference on a 50,000-word project is real money. Content marketers pay per word. In all these cases the rule is the same: find out which counter the recipient uses, and write to that one.
Which number do you trust?
Three practical rules:
- Match the venue. Submitting to a portal? Paste into it (or into Word, the most common reference) and trust that number.
- Be consistent. Track drafts with one tool so day-to-day comparisons are apples to apples — the absolute number matters less than the trend.
- Leave margin. If the limit is 500 and counters disagree by ~2%, aim for 490 in your own tool. Margin is cheaper than a rejected submission.
Our counter's definition, stated plainly
Every counter should state its rule, so here's ours: the EduTray Word Counter splits on whitespace — each maximal run of non-whitespace characters is one word. Hyphenated compounds count as one word, standalone dashes count as tokens only if your text puts spaces around them (we count them, Word doesn't — now you know why we might differ by a hair), and "1,200" is one word. Sentences split on full stops, exclamation marks and question marks, paragraphs on blank lines. It's a simple, predictable rule — and now that you know it, you'll never be surprised by it.